磁效应回转角度测量的GRNN算法研究  

Research on rotation angle measurement using magnetic effects based on GRNN algorithm

作　　者：陈时毅 卢泽洵 廖普 刘文文[1] 胡鹏浩[1] CHEN Shiyi, LU Zexun, LIAO Pu, LIU Wenwen, HU Penghao(School of Instrument Science and Opto-electronics Engineering, Hefei University of Technology, Hefei 230009, China)

机构地区：[1]合肥工业大学仪器科学与光电工程学院,安徽合肥230009

出　　处：《中国测试》2018年第9期102-108,共7页China Measurement & Test

基　　金：国家自然科学基金项目(51475133;51475132)

摘　　要：针对等效磁荷法在反解智能球铰链回转角度中存在的不足,该文尝试采用神经网络算法来计算、仿真和获取精密球铰链回转中的二维空间转角。根据仿真结果得出最优算法,给出提高测量精度的可行性方案。对神经网络GRNN算法进行验证,并与等效磁荷模型法的计算精度进行比较。搭建测量实验平台,实验结果与前期研究结果相比,在±10°范围内角度平均测量误差降低0.52′,在±20°范围内误差降低12.44′,表明测角精度得到改善,且随着量程的增大,GRNN算法的角度反解误差在减小,同时该算法计算精度不依赖于球铰链各项结构参数,为进一步提高球铰链空间回转角度测量精度提供新的可能。Aiming at disadvantages of equivalent magnetic charge model in rotate angle solution for intelligent ball hinge, neural network algorithms are selected to calculate, simulate and obtain the spatial angle. A suitable algorithm is determined according to the simulation results, and a feasibility scheme is proposed to improve the measurement accuracy. The effectiveness of neural network GRNN algorithm is proved after comparing with the calculation precision of equivalent magnetic charge model method. A measurement and test platform is set up and the test results show（comparing with preliminary research results） that the GRNN algorithm make the average error decrease 0.52′ within ±10° measuring range and reduced the average error as much as 12.44′ in±20° measuring range. This indicates that not only the angle measurement accuracy is improved, but also the measurement accuracy increases in stead of decrease with the measuring range enlarging. Meanwhile, the accuracy of GRNN does not depend on the structural parameters of the ball hinge. The exhibited characteristics of GRNN algorithm paves the way for further improving the measurement accuracy of intelligent ball hinge in the near future.

关 键 词：球铰链 空间转角测量 GRNN算法 磁效应 

分 类 号：TH712[机械工程—测试计量技术及仪器]